Ingest (Vector)
Lightweight log shipping and transformation pipeline. Handles syslog and one Windows host lane with minimal CPU footprint.
v1.0.0 • local-first • homelab-ready
A local-first security detection playground for developers and homelab users. Run it with Docker Compose, click one button, and watch Hayabusa detect suspicious login activity on your own machine.
Detection triggered: Failed login burst on a local machine
$ hayabusa --smoke-test
[OK] Vector ingest lane: Active
[OK] ClickHouse: Writing 6.42 MB/s
[SQL] SOURCE = security.events; WINDOW = 5m
{ "timestamp":"2026-04-09T05:42:18Z", "src_ip":"192.168.1.45", "user":"admin" }
{ "reason":"failed_password", "count":6, "burst_window":"2m" }
!!! ALERT: BRUTE_FORCE_DETECTED [source: 192.168.1.45]
Operational logic
Lightweight log shipping and transformation pipeline. Handles syslog and one Windows host lane with minimal CPU footprint.
Rules evaluate failed-logon patterns every 30 seconds and write matches into
security.alert_candidates.
Native Grafana webhook routing turns detections into a real payload the local alert sink can receive and log.
Five steps, one story.
Telemetry streams from syslog or Windows forward into Hayabusa.
Normalized events land in ClickHouse for rapid lookup.
Scheduled SQL checks identify suspicious login bursts.
Matches are written to alert_candidates for alerting and review.
Grafana posts a real payload and alert-sink confirms delivery.
What Hayabusa detects today
Small enough to explain, real enough to demo.
A complete ingest → detect → alert pipeline using simple, composable components.
Live detection pipeline
Live detection pipeline
Real demo capture coming soon.
See a real login attack get detected on your machine.
Built for developers, self-hosters, and homelab users who want immediate local feedback without cloud dependencies or enterprise overhead.